MARIE DE ZÉLICOURT, LAURENT BUTEAU, FRANCIS FAGNANI & PIERRE JALLON

Seizure 2000; 9: 88 95 doi: 10.1053/seiz.1999.0364, available online at http://www.idealibrary.com on The contributing factors to medical cost of epilepsy: an estimation based on a French prospective cohort study of patients with newly diagnosed epileptic seizures (the CAROLE study) MARIE DE ZÉLICOURT, LAURENT BUTEAU, FRANCIS FAGNANI & PIERRE JALLON Cemka; Hôpital Cantonal de Genève, Unité d Épileptologie Clinique et d Electroence Phalographie Correspondence to: Docteur Marie de Zélicourt, CEMKA-43, Boulevard Maréchal Joffre, F-92340 Bourg-la-Reine, France Epilepsy is a common neurological condition with significant resource implications. An estimation was performed in France of the direct medical cost of patients presented with newly diagnosed seizures and followed during the first two years after diagnosis. This estimation was based on the service utilization data collected from French prospective cohort study (CAROLE: 1942 patients enrolled). Costs were estimated in a societal perspective in 1998 value. The impact on the costs of different factors like age, aetiologic categorization and severity of seizures (type and number of seizures), and treatment by anti-epileptic drugs (AEDs) was analysed. The mean annual direct epilepsy-related costs per patient were estimated to be 14 305 F and 3766 F for the first and the second year of follow-up respectively, 68% and 40% of the costs were devoted to inpatient care. Costs during the first year were highly sensitive to aetiologic categorization of seizures at inclusion and to other clinical parameters. Secondyear costs had a much lower variance and were sensitive to frequency of seizures and the fact of being treated or not by AEDs. Our data emphasize the importance of seizure control as means of reducing the costs of epilepsy especially during the first year of follow-up. c 2000 BEA Trading Ltd Key words: epilepsy; cost of illness; seizures; direct costs. INTRODUCTION Epilepsy is an important public health problem because of its burden on individuals and society in terms of morbidity, direct and indirect costs. Epileptic seizures are common events. According to epidemiological surveys performed in industrialized countries, the prevalence of epilepsy is estimated to be approximately 5 8 per 1000 with an incidence of approximately 50 100 per 100 000 per year 1. The medical and social care of this disorder entails large costs related to hospitalization, diagnostic procedures, drugs and social assistance 2, 3. It is generally accepted that the direct medical costs of diagnosed epilepsy are very dependent on the time period considered after diagnosis 4, 5. The first year integrates all costs related to diagnosis for all incident epileptic cases, whereas in the subsequent years a large proportion of patients will never experience any relapse, thus reducing the mean costs. Large differences then depend on the prognostic groups of patients 5. The cost of epilepsy has been rarely estimated on the basis of direct observations of real medical consumption on large samples of patients because of the scarcity of large population studies in this area 4, 6, 7. It was then an interesting opportunity to derive such cost estimates from a cohort study of epileptic patients followed for 2 years after initial diagnosis. This French cohort study named CAROLE (Active Coordination of the Longitudinal Observational Network in Epilepsy) was characterized by a large sample size and the collection of very detailed clinical parameters on the enrolled patients that allowed an analysis of the main possible factors of variation. However, the variables available to describe non-medical direct costs and indirect costs were considered inadequate to permit any reliable estimation despite the fact that indirect costs represent a major part of the societal burden of epilepsy 2, 5, 8. Neither was any quality-of-life questionnaire collected 9, 10. Therefore, the results presented here are restricted to direct medical costs. 1059 1311/00/020088 + 08 $35.00/0 c 2000 BEA Trading Ltd

The cost of epilepsy in France 89 MATERIALS AND METHOD Study population The CAROLE study is a prospective, observational, and naturalistic survey of patients with newly diagnosed unprovoked seizures, enrolled by volunteer specialists in neurology and neuropediatry throughout France. Participating specialists (N = 243) were asked to enroll all consecutive different patients seen in their current public and/or liberal practice from 1 May, 1995 until 30 June, 1996. One thousand nine hundred forty two patients could be enrolled and followed up for 2 years. The cohort included men and women aged more than 1 month, presenting with a newly diagnosed unprovoked seizure. The only exclusion criteria concerned the patients inability to participate in a long-term follow-up for social reasons. Data collection Data was collected and recorded on standard case report forms at inclusion (baseline) and at each patient s follow-up visit for a minimum of 2 years. Being designed as a purely observational and naturalistic survey, the follow-up was to interfere as little as possible with the patient s usual care. In particular, no attempt has been made in the present analysis of the data to perform a critical review of diagnosis and care 11. Consequently, no specific visit or procedure was planned in the study protocol and hence we considered that there were no protocol-induced costs. In case of no visit occurring on a 6-month period, investigators were only asked to have an informal contact with the patient (or his GP) to collect minimal information about his current clinical status. All documents were checked in real time by the medical monitor. Anonymity of patients was strictly maintained. At inclusion, data collected addressed the description of the first diagnosed seizure (index seizure) in terms of circumstances, duration and type, and possible history of other seizures. General characteristics on health, clinical examination, procedures, drug therapy, other use of medical services, hospitalizations, and socioeconomic status of the patients (or their relatives for the younger patients), diagnostic conclusions and therapeutic decisions were also recorded. At each follow-up visit, physicians were asked to collect information about the recurrent seizure(s), use of medical services, and modification of socioeconomic status since the last contact. They also had to record results of recent diagnostic procedures, to re-evaluate their diagnosis, and to describe current treatments (drug compliance, side effects, etc.). Direct costs estimation The general approach used here is a so-called incidence-based cost analysis restricted to a 2-year follow-up period after diagnosis 5. This approach is based on a longitudinal analysis of individuals taking into account the temporal aspects of epilepsy. This method is an alternative to cross-sectional approaches based on observation of samples of patients present in a population at all stages of evolution of the disease. Direct medical costs included all epilepsy-related items of medical consumption, i.e. physicians visits, hospitalizations, diagnostic investigations, laboratory tests, and antiepileptic drugs. Non-medical costs (transportation costs, residential care and unpaid care or services provided by patients relatives) were not documented in our study. Costs were estimated for each patient, using the items provided in the case report forms in combination with unit cost determined from external sources. No database belonging to National Sickness Funds or other third-party payers was used to estimate individual costs. A societal perspective was chosen, meaning that all costs either reimbursed by the public Sickness Funds and/or private insurance or patient s out-ofpocket expenses were included. Costs of drugs were calculated on the basis of the current prices used in France. They are restricted to outpatient consumption. It may be noticed that in France all drugs prescribed for outpatient care have a regulatory national price resulting from an agreement with Health Authorities. These prices are listed in the VIDAL 98. Costs are then calculated by using these unit prices, dose and duration of prescriptions. All items of medical consumption used for inpatient care were valued globally by multiplying a daily cost per category of ward by the corresponding length of stay. Main unit costs used and their sources are summarized in Table 1. Concerning outpatient care, unit costs are based on reimbursement rates currently in use in France. All costs have been converted to French Francs 1998 based on evolution rates of health expenditures. Finally, direct medical costs were calculated as the sum of all costs described here above for each patient on a per annum basis (i.e. year 1 and year 2 after diagnosis). Statistical methods Patients were shared according to age group (less than 16 years, 16 60 years and over 60 years), aetiologic categorization at inclusion (idiopathic, cryptogenic, symptomatic) and seizure severity after 1- year follow-up (mean number of seizures per month).

90 M. de Zélicourt et al. Table 1: Main unit costs and sources of data (1FF = 0.15 Type of costs Unit cost Source (FF 1998) Inpatient care: Analytical Accounting from Assistance Emergency room 2701 Publique Hôpitaux de Paris taken as Neurology ward 3061 reference prices. : GPs visits 115 Neurologist and other visits 150 Nomenclature Générale des Actes Usual EEG 411 Professionnels CT scan (contrast media not included) 900 MRI (contrast media not included) 2300 Table 2: Socio-demographic characteristics of the cohort. Age group 1 month 15 years > 15 60 years >60 years Gender, females (%) 49 44 47 Age, mean (SD) 7 (4) 31 (12) 70 (7) Working, of the active population (%) 76.2 Number of patients (%) With 1-year follow-up 717 (40.8) 870 (49.5) 169 (9.7) With 2-year follow-up 633 (43.2) 696 (47.6) 135 (9.2) Cumulative length of follow-up 1350 1566 304 (patient years) Since cost variables were not normally distributed, we used the non-parametric test for global cost differences across patient categories (Kruskall Wallis) and, in case of significance, Mann Whitney tests were used for category-by-category comparisons. Finally, a multivariate model of analysis of variance was used in order to quantify the individual contribution of each clinical factor according to a descending hierarchical approach. This method allows analysis of the relationships between the annual direct costs considered as a quantitative variable and a series of clinical and demographic explanatory variables described in categories (2 or 3 modalities). The analysis was performed in using the PROC/GLM from SAS R12. RESULTS The CAROLE cohort The main characteristics of the cohort are presented in Table 2 to allow comparison with other available epidemiological data 4, 5. The cumulative length of follow-up was 3220 patient years. Of 1942 patients initially enrolled, 1756 (90.4%) and 1464 patients (75.4%) respectively had a 1- and 2-year followup with a fully documented medical consumption. The mean values presented below were calculated for the first year on patients with at least a fully documented 1-year follow-up period. The same principle was applied for the second year. The younger age group (1 month to 15 years) represented 40.8% of the cohort and the older group (over 60 years) 9.7%. The percentage of the adult population at work was 76.2%. Health care management Table 3 presents the rates of utilization for different services, procedures and treatment related to epileptic seizures. The bottom line takes into account at least one of the whole set of medical items. For the first year, the result is of course 100% due to the design of the study, but decreases to 87 90% on the second year, meaning that 10 13% of patients had no medical consumption in relation with epilepsy in the second year of follow-up. Concerning inpatient care, the percentage applies to any hospital stay, whatever its duration, including day hospitalizations. Concerning the utilization rates of diagnostic procedures, the figures presented in Table 3 apply only to those performed in out-patient care. The perspective was here to estimate costs and not to fully describe the medical management of these patients. This aspect will be described in an upcoming publication. The percentage of patients in each age group treated by at least one AED was calculated whatever the duration of treatment during the

The cost of epilepsy in France 91 year considered. No minimum duration was used as a threshold. Annual direct medical cost of epilepsy The results on annual direct costs of patients with newly diagnosed seizures are presented in two steps in order to avoid any misinterpretation of the data. In Table 4, the mean annual costs per patient in each age group having used the corresponding item of medical care are displayed. For example, the mean cost of inpatient care is 11 335FF (1FF = 0.15 = 0.17US$) during the first year of follow-up for a patient 16 60 of age, who has been hospitalized. Table 3 shows that the corresponding mean length of stay was 5 days and that only 66% of patients were concerned. Table 5 supplies the corresponding costs calculated on the whole population in each age group, whether hospitalized or not. This last value is used for global estimates of mean costs. The results suggest that direct medical costs for these patients are depending on the age group. But, these differences are only statistically significant for the first year of follow-up (P = 0.0001) and this is true only for the older age group (over 60 years) as compared with the two others. The most significant contributor to the total cost was inpatient care regardless of age. It accounted for 68% in the first year of follow-up and for 40%, i.e. at a same proportion as AEDs, in the second year. Direct costs according to clinical factors The preceding results were disaggregated according to the main clinical parameters available in the CAROLE study. These clinical criteria were defined as following: Etiologic categorization : of patients at inclusion according to the ILAE recommendations (idiopathic, cryptogenic and symptomatic) 13 ; Seizures pattern : this item concerned the seizures experienced by the patient at the time of diagnosis (index seizure) or the history of seizures having occurred previously, if any. Three categories were used: single : this category concerns the patients presenting with only a single seizure at the time of diagnosis (one seizure at the time of inclusion); multiple : this category concerns the patients with an history of undiagnosed seizure(s) before diagnosis whatever the time elapsed until the index seizure; inaugural status epilepticus : in this last category, we included also the cases of inaugural multiple seizures (cluster of seizures) having occurred in a single 24-h interval. Seizure frequency at 1 year 14. It was estimated as a monthly mean calculated over a 6-month period about 1 year after diagnosis. The distribution of this frequency is very skewed and we chose to put a cut-off value at 30 in the analysis among those presenting with an active epilepsy. Treatment by AEDs. It was considered for each year of follow-up separately and used here as a dichotomous variable. As mentioned earlier, no minimum treatment duration was used to categorize a patient as being treated. The results presented in Table 6 are ranked for each parameter in ascending order in terms of mean costs for the first year of follow-up. In each variable, the global differences in costs per category appeared statistically significant ( P < 0.0001). However, the differences in costs for the second year were not significant in two cases: when comparing cryptogenic vs. idiopathic seizures, and multiple seizures vs. inaugural status epilepticus. It may be noticed that the initial characteristics of seizures (aside from the etiologic class) do not modify significantly the second year of follow-up costs. The interpretation of these last results was, however, made difficult due to the existing correlations among all parameters. For this reason, a multivariate model of variance was used, the results of which are presented in Table 7. A reference subgroup of patients was defined as presenting with the less-severe condition among the largest age group. It was comprised of adults, with a single cryptogenic seizure at diagnosis followed by no recurrent seizure at 1-year follow-up and who were never treated by AEDs. The mean annual cost in this subgroup for the first year of follow-up was found to be 7664F (value of the intercept coefficient). The model provides parameter estimates that may be interpreted as the extra cost contributions to this baseline value associated with the category considered, all other parameters being controlled. The P-values in the second column of Table 7 indicate the statistical significance of each variable in the model (difference from zero). The results are presented in Table 7. In summary, being in one among the following categories respectively, brings an extra cost in the order of magnitude of 4000 5000FF: either being younger or older (under 15 or over 60 years of age), or having an active epilepsy (with a number of seizures at 1 year of less

92 M. de Zélicourt et al. Table 3: Drugs and medical service utilization rates (to nearest 1%) by age group and by year. Age Group 1 month 15 years >15 60 years >60 years (% of patients) (% of patients) (% of patients) Year 1 Inpatient care 63 (6,3 a ) 66 (5,0 a ) 73 (10,7 a ) Neurologist visits 97 (3,6 b ) 97 (3,5 b ) 94 (3,1 b ) Other physician visits 86 (6,4 b ) 83 (6,1 b ) 86 (7,6 b ) EEG, at least 1 91 81 64 CT scan or MRI, at least 1 39 57 44 Laboratory tests 53 51 53 AEDs 81 81 86 Other investigations 3 2 5 At least one medical item 100% 100% 100% Year 2 Inpatient care 12 (7,7 a ) 6 (5,2 a ) 7 (8,8 a ) Neurologist visits 77 (1,6 b ) 63 (1,4 b ) 56 (1,8 b ) Other physician visits 25 (2,4 b ) 26 (3,1 b ) 26 (7,8 b ) EEG, at least 1 59 39 28 CT scan or MRI, at least 1 3 3 2 Laboratory tests 28 23 33 AEDs 81 82 85 Other investigations 1 0 0 At least one medical item 90 87 88 a Average annual length (number of days) of hospitalization by inpatient. b Average number of visits by patient using this medical service. than 30 per month) or being treated by AEDs. Conversely, a patient presenting with idiopathic seizures decreases the baseline cost by 5310F. On the other hand, a patient presenting with symptomatic seizures or suffering more than 30 seizures per month at 1 year, contributed much more to the baseline cost (9045F and 28 936F, respectively). The first-year period costs appeared highly sensitive to all clinical parameters considered, but with different contributions. These could be sorted in descending order: (1) aetiologic categorization at inclusion (idiopathic, cryptogenic, symptomatic); (2) number of seizures at 1 year; (3) age group (15 60, under 15, over 60 years); (4) pattern of seizure ( single, multiple, inaugural status epilepticus ); (5) being treated or not by AEDs. Conversely, the second-year costs had a much lower variance and were only sensitive to two characteristics of the patients: the disease severity (number of seizures) at one year and the fact of being treated or not by AEDs. DISCUSSION The mean annual direct medical epilepsy-related costs per patient were estimated 14 305F and 3766F for the first and the second year of follow-up, respectively, with inpatient care services accounting for 68% and 40% of the total mean annual direct costs (see Table 5). This estimation was based on the service utilization data collected in a cohort study in which 1942 patients were enrolled. Costs were estimated in a societal perspective in 1998 value. It is difficult to compare these results with similar published values due to the differences in perspective and methodology among studies. Concerning French data, we have not found any comparable results in the literature. Another study which provided direct costs for the period after diagnosis was performed in the UK and based on the NGPSE 4. Our results were much higher: 611 (about 6110FF) vs. 14 305FF for the mean direct costs for the first year of follow-up, that is more than two times higher. The ratio between the costs of the first and the subsequent years are, however, similar (36.2% vs. 30% in our study) and also the share of costs devoted to AEDs out of the total direct costs (12% and 8% for the CAROLE study in the first year). It would be interesting to know if this difference was due to discrepancies in terms of unit costs or medical resource consumption. It was difficult to fully answer this question in absence of detailed unit cost data and direct observations on real utilization rates in the UK study. It is probable that both aspects play a role. It is unclear whether the UK study included all diagnostic procedures performed at the moment of enrolment as

The cost of epilepsy in France 93 Table 4: Mean cost/user of each item by age group and by year (1FF = 0.15 Type of costs Age group Total 1 month 15 years >15 60 years >60 years Year 1 (mean) Inpatient care 17 689FF 11 335FF 21 003FF 14 870FF Neurologist visits 544FF 501FF 462FF 515FF Other visits 292FF 302FF 380FF 306FF EEG 1 634FF 1 053FF 868FF 1 298FF CT scan or MRI 2 847FF 2 821FF 2 576FF 2 808FF Laboratory tests 492FF 484FF 495FF 488FF AEDs 1 159FF 1 596FF 1 506FF 1 410FF Other investigations 383FF 400FF 504FF 410FF Total Direct Costs 15 643FF 12 212FF 19 404FF 14 305FF Year 2 (mean) Inpatient care 20 246FF 10 658FF 19 965FF 17 179FF Neurologist visits 268FF 238FF 271FF 255FF Other visits 392FF 363FF 972FF 432FF EEG 776FF 523FF 546FF 663FF CT scan or MRI 2 679FF 2 906FF 2 363FF 2 767FF Laboratory tests 315FF 301FF 338FF 312FF AEDs 1 615FF 2 064FF 1 699FF 1 837FF Other investigations 163FF 199FF 168FF Total Direct Costs 5 286FF 3 375FF 3 953FF 4 270FF Table 5: Mean direct costs per patient included in the cohort by age group and by year (in FF 1998) (1FF = 0.15 Type of costs Age group Total % 1 month 15 years >15 60 years >60 years Year 1 (mean) Inpatient care 11 102FF 7 479FF 15 410FF 9 722FF 68 Neurologist visits 530FF 484FF 434FF 498FF 3 Other physician visits 252FF 252FF 328FF 259FF 2 EEG 1 436FF 837FF 509FF 1 050FF 7 CT scan or MRI 1 112FF 1 605FF 1 143FF 1 359FF 10 Laboratory tests 261FF 248FF 263FF 255FF 2 AEDs 939FF 1 297FF 1 292FF 1 150FF 8 Other investigations 11FF 10FF 24FF 11FF 0 Total Direct Costs 15 643FF 12 212FF 19 404FF 14 305FF 100 Year 2 (mean) Inpatient care 2 527FF 628FF 1 331FF 1 514FF 40 Neurologist visits 206FF 151FF 150FF 175FF 5 Other physician visits 97FF 94FF 252FF 110FF 3 EEG 433FF 192FF 142FF 292FF 8 CT scan or MRI 89FF 92FF 53FF 87FF 2 Laboratory tests 88FF 70FF 110FF 81FF 2 AEDs 1 312FF 1 696FF 1 447FF 1 507FF 40 Other investigations 2FF 0FF 0FF 1FF 0 Total Direct Costs 4 751FF 2 924FF 3 484FF 3 766FF 100 the French study did. The hospitalization rate was thus 11% in UK as compared with 65% in France with an average length of stay of 4.5 and 6 days respectively. But most hospitalizations in France were caused in the first year by the index seizure(s) care and associated work-up. The referrals to neurologists are also probably very different between the two countries due to the large differences in the number of these specialists. Another explanation may be related to the distribution by age of our population of newly diagnosed patients. As compared with other epidemiological data in the literature 4, 5 it appears that the CAROLE sample is probably under-represented in the over 60 years age group. The proportion of these patients was around

94 M. de Zélicourt et al. Table 6: Mean annual direct costs of epilepsy per patient in the cohort according to different clinical categories (in FF 1998) (1FF = 0.15 Year 1 Year 2 Etiologic categorization at inclusion a Symptomatic 25 676FF 5 797FF Idiopathic Cryptogenic 9 036FF 13 539FF 2 580FF 3 775FF Pattern of seizure(s) at inclusion a Single 12 946FF 2 563FF Multiple b 14 213FF 4 609FF Inaugural status epilepticus b 20 783FF 3 665FF Seizure frequency at 1 year a None 12 555FF 2 455FF 1 29 per month 17 196FF 4 653FF >30 per month 47 099FF 37 588FF Treated by AEDs a No Yes 9 410FF 15 408FF 4 636FF 451FF Total 14 305FF 3 766FF a P < 0.0001 except idiopathic vs. cryptogenic and multiple vs. inaugural status epilepticus for the second year (non significant difference). b See definition in the text. Table 7: Model of analysis of variance of direct medical cost during the first year of follow-up (1FF = 0.15 Variable P Modalities Reference Parameter P group estimates (FF) Age 0.0019 1 month 15 years 3 547 0.0026 >15 60 years x 0 >60 years 4 715 0.0126 Aetiologic category 0.0001 Idiopathic 5 310 0.0001 Cryptogenic x 0 Symptomatic 9 045 0.0001 Pattern of seizures at inclusion 0.0066 Not status epilepticus a x 0 Status epilepticus a 5 142 0.0066 Number of seizures per month at 1 year 0.0001 No seizure x 0 1 30 per month 3 690 0.0049 30 per month 28 936 0.0001 Treated by AED 0.0112 Yes 3 529 0.0112 No x 0 Intercept 7 664 0.0001 a Inaugural (see definition in the text). 10% here while it represented 24% in UK 4 and 33% in the US 5. However, correcting for this potential bias will yield an increase in the mean costs in France. CONCLUSIONS Our data emphasize the importance of seizure control as a means of reducing the costs of epilepsy, especially during the first year of follow-up. ACKNOWLEDGEMENT This study was supported in part by an unrestricted grant by Parke-Davis France. REFERENCES 1. Sander, J. W. A. and Shorvon, S. D. Incidence and prevalence studies in epilepsy and their methodological problems; a review. Journal of Neurology, Neurosurgery and Psychiatry 1987; 50: 829 839. 2. Pachlatko, C. The cost of epilepsy care to the community (editorial). Seizure 1997; 6: 415 417. 3. Griffiths, R. I. et al. Payer costs of patients diagnosed with epilepsy. Epilepsia 1999; 40: 351 358. 4. Cockerell, O. C. et al. The cost of epilepsy in the United Kingdom: an estimation based on the results of two populationbased studies. Epilepsy Research 1994; 18: 249 260. 5. Begley, C. E. et al. Cost of epilepsy in the United States: a model based on incidence and prognosis. Epilepsia 1994; 35: 1230 1243. 6. Beran, R. G. and Pachlatko, C. Report of the International League Against Epilepsy Commission on economic aspects of epilepsy. Epilepsia 1996; 37: 506 508. 7. Beran, R. G. and Pachlatko, C. Final report of the ILAE Commission on Economic Aspects of Epilepsy, 1994 1997. International League Against Epilepsy. Epilepsia 1997; 38: 1359

The cost of epilepsy in France 95 1362. 8. Loiseau, P. Epilepsy: epidemiology, economic data. Journal of Neuroradiology 1983; 10: 119 121. 9. Gilliam, F. et al. Patient-validated content of epilepsy-specific quality-of-life measurement. Epilepsia 1997; 38: 233 236. 10. Meador, K. J. Research use of the new quality-of-life in epilepsy inventory. Epilepsia 1993; 34: S34 S38. 11. Nowack, W. J. Epilepsy: a costly misdiagnosis. Clinical Electroencephalography 1997; 28: 225 228. 12. Slinder, B. K. and Glantz, S. A. Multiple linear regression is a useful alternative to traditional analyses of variance. American Journal of Physiology 1988; 255: R353 R367. 13. Commission on Epidemiology and Prognosis, International League against Epilepsy. Guidelines for epidemiologic studies on epilepsy. Epilepsia 1993; 34: 592 596. 14. Baker, G. A. et al. Current issues in the management of epilepsy: the impact of frequent seizures on cost of illness, quality of life, and mortality. Epilepsia 1997; 38: S1 S8.